Back to search

Aristotle University Of Thessaloniki (AUTH)

Information technology and computational intelligence methods for participatory environmental sensing applications

Abstract

dc:description

Air pollution is a major problem, directly affecting people’s health and quality of life. The degradation of the atmospheric environment (AE) due to increased levels of air pollution is considered as an existing public health threat in urban and rural areas. In 2016, it was estimated that outdoor air pollution was responsible for 4.2 million premature deaths worldwide per year (WHO, 2018). One aspect that can mitigate the negative impact of the poor quality of the AE in human health is to increase citizens’ awareness of the problem, in a timely, location-based, personalised and proactive manner. In other words, it is important to have accurately informed citizens about the existing air quality (AQ) of their area of interest, focusing on their personals sensitivities and needs. In the era of rapid technological change and information explosion, the ultimate goal is to take advantage of the availability of data and turn them into useful, targeted knowledge for both practitioners and citizens. Being able to “interpret” and “understand” the massive collection of data automatically, may lead to innovative quality of life (QoL) services. Within the context of the dissertation, our research efforts cover a very large part of the “life-cycle” of AQ data, in all its expressions, involving processes for their primary production, collection, processing, modelling and transformation, up to the semantical representation and enrichment of the derived content. We target to deconstruct the concept of personalised environmental information and decision support system, by studying and contributing in each of its relevant aspects, i.e., (a) from analysing and modelling heterogeneous content related to the quality and the impact of the AE to human health, to forecasting future trends and unravelling hidden relationships among the examined/combined parameters, and (b) from structuring the knowledge gained through the analysis, enriched by the knowledge provided by environmental and health domain experts, to designing and performing reasoning for personalised recommendation provision to sensitive health groups for supporting their daily choices and preserving their QoL. Our target can be considered as an interdisciplinary task which regards the integration of different research fields and methods for achieving separate research objectives, with different specifications, requirements and tools. We bring together Computational Intelligence (CI), Data mining (DM) methods and Semantic Web technologies (Ontologies). More specifically, classification (Artificial Neural Networks, Decision Trees) and clustering (k-means, Self-Organising Maps) techniques are utilised for modelling the AQ domain in different areas of interest, and for forecasting specific AQ-related parameters (short/long-term pollutants’ concentrations, AQIs, ratios, and more). The applicability of unsupervised learning is also studied for revealing hidden relations between the examined parameters, i.e., the similarity of the AQ behaviour in neighbouring areas, the positive/negative relations between air pollutants, pollen and weather data, etc. Furthermore, with the use of Natural Language Processing (NLP) and clustering techniques, we study and aim to bring out for the first time the relevance of official AQ related measurements (hard sensors) with user-generated content reported in social media, web 2.0 or dedicated platforms (soft sensors), which involve reports and personal estimations about the impact of the AE quality to people’s health. Taking advantage of the semantic expressiveness and structural representation capabilities of ontologies, we present how Semantic Technologies can be applied in order to formalise a structural framework that integrates the involved parameters for both AQ-related information and recommendation provision. Information about the existing AQ conditions, user-profile details, and standards and relations that frame the impact of AQ to human health, as those were suggested by AQ/health experts, are expressed in the form of concepts, semantic relations and rules. The proposed representation model (schema) and rule-based reasoning mechanism is modular and extensible to new sources and needs; additional AE-related elements (air pollutants, pollen, etc.), relevant types of event definitions (AQ episodes, severity levels of the impact of AQ to human health, etc.), multiple recommendations per use case can be easily supported by adopting and extending correspondingly the suggested representation, thus preserving the main principle of ontologies which is the reuse of their structured content. Aforementioned research outcomes were tested with the aid of real world applications via relevant research projects. Based on the conducted research and the derived results, we consider the proposed approach for analysis, modelling, and extraction and dissemination of personalised, enriched information, as successful. The methodologies that have been adopted and adjusted to the needs of the different research tasks have shown the advantages of: (i) the CI methods in modelling heterogeneous data and for handling different complex tasks in forecasting, representing and extracting new knowledge, and of (ii) the Semantic Web technologies in reasoning and inferencing personalised information, targeted to the needs of the end-users. Below, we summarise the most important contributions of our research:* The proper combination (hybrid approach) and parameterisation of well-known DM and CI methods for different classification/clustering tasks and for describing the different domains of interest, on the basis of analysing different hard/soft-sensor data.* The adoption of a cluster-based solution for the extraction of hidden information and relations that exist in user-generated content; the proposed solution can efficiently analyse natural-language documents, by taking into account the semantic relations of their content.* The proof of concept that soft- and hard-sensor data describe similar conditions and events, which was examined and proved, on the basis of different sources of data and different domains of interest within the scope of AE and QoL. * The creation of a prototype framework for formalising and structuring both the sourced and the described knowledge, in order to represent and disseminate timely information related to AQ status as well as targeted recommendations on the basis of the impact that the existing AQ might have to citizens’ health. The ultimate advantage of the proposed solution is that it is domain agnostic and can be adopted in different domains of interest and application, following a similar mechanism for defining and representing the different user-profiles and user-profile driven recommendations per case. * The creation of a hybrid reasoning process, on the basis of ontology-based representation of concepts and rules that bind the domain of interest, where simple axioms ensure the consistency and validity of the declared relations (hierarchy, domain, range, etc.), while complex rules handle multifaceted reasoning tasks that concern the provision of proper, personalised, health-related suggestions, targeting at specific sensitive health groups, as proposed by the AQ/health experts.

Degree

thesis:*
Grantor dc:publisher
Aristotle University Of Thessaloniki (AUTH)
Year dc:date
2020

Author and committee

dc:creator, dc:contributor.*
Authors dc:creator
  • Riga, Marina
  • Ρήγα, Μαρίνα

Subjects

dc:subject × 32

Rights

Language dc:language
eng

Identifiers

dc:identifier.*
Identifier
10.12681/eadd/47487

Chain of custody

source
Harvested from
Greek National Archive of PhD Theses
Base URL
phdtheses.ekt.gr/eadd_oai/request
Last updated
2026-11-18
Source record
OAI-PMH GetRecord
citation

Riga, Marina; Ρήγα, Μαρίνα. Information technology and computational intelligence methods for participatory environmental sensing applications. Aristotle University Of Thessaloniki (AUTH), 2020. http://hdl.handle.net/10442/hedi/47487